Coding Agent Prototype
A code-oriented LLM agent with task parsing, context management, tool calling, command execution, verification, and error repair.
This project builds a coding-agent prototype around a large language model. The system is designed to understand code tasks, gather repository context, edit files, execute commands, inspect results, and iterate on failures.
The core workflow includes task parsing, context management, tool calling, command execution, result feedback, code understanding, file modification, validation, and debugging loops.